Automated function prediction currently takes one of four general approaches:

• evolutionary: making use of conservation of sequence or structure

• comparative genomics: identification of gene fusion, phylogenetic profiles, and conservation of gene order

• cellular: use a large proteomic database to define networks of protein interactions

  Interact means synthesize a substrate for or bind to, creating a functional complex

• metabolic: use a large database to define networks of biochemical pathways

We consider an example or two from each of these categories.

Evolutionary - The Rosetta Stone

• Some pairs of interacting proteins have homologs in another organism fused into a single protein

• For example, GyrA (804 aa) and GyrB (875 aa) subunits of E. coli DNA gyrase have strong similarities to portions of the topoisomerase II of yeast (1429 aa)

• This similarity might have been used to predict that GyrA and GyrB interact in E. coli
  

• Eisenberg et al. (Science, 1999, 285, 751) designate the fused protein (gene) as a Rosetta Stone

• Searching the E. coli genome they found 4290 pairs of sequences with significant similarity to a single protein in another genome

• Whole pathways might be reconstructed: A interacts with B, B interacts with C, and so on

• This was demonstrated for several pathways in E. coli:

  Inferred Pathways in E. coli
  A = shikimate; B = purines; C, D = predicted from A, B

  Each symbol represents an enzyme in the given pathway: for example, AroB is dehydroquinate synthase (3clh); AroD is dehydroquinate dehydratase (3JS3); PurK is a phosphoribosylaminoimidazone carboxylase....

Genomics - Inferring Genes Belonging to the Same Operon

Operon organization in prokaryotes results in a series of adjacent genes being transcribed into a single mRNA, which codes for the synthesis of multiple proteins

• The proteins from operons usually have related functions, form common complexes, or participate in shared biochemical pathways

• If genes can be established as belonging to an operon, and any of the resulting proteins have known function, inferring the other functions is possible

• Existence of an operon can be inferred from intergenic distances:

  Possible Gene Arrangements
  RBS = ribosome binding site; UTR = untranslated region

• Only Case 4 represents an operon

• What base pair distance represents a reasonable threshhold? Eisenberg et al. (Genome Biology, 2003, 4, R59) used the genome of M. tuberculosis, and attempted to link unannotated genes to annotated ones:

  Threshhold (bp)	Non-Annotated genes with links	Non-annotated linked to one or more annotated
  0	474	217 (14%)
  25	786	412 (27%)
  50	913	521 (34%)

Genomics - Phylogenetic Profiles

The PLEX database stores:

• gene sequences

• chromsomal positions

• precomputed phylogenetic profiles

• Rosetta Stones

and can be searched iteratively using either a sequence or a phylogenetic profile. It provides links for gene neighbors and Rosetta stones.

I submitted the sequence for a fungal quinone oxidoreductase modeled by our group:

MCFPSKRRKD GSPEEGGRIK RSRSAQEPAE STNTPAPPTS TGTKPTTTQT
TDTTMSSPRL AIVIYTMYGH VAKLAEAIKS GIEGAGGNAS IFQVAETLSP
EILNLVKAPP KPDYPVMDPL DLKNYDGFLF GIPTRYGNFP VQWKAFWDST
GPLWASTALC GKYAGLFVST GSPGGGQEST LMAAMSTLVH HGVIYVPLGY
KYTFAQLANL TEVRGGSPWG AGTFANSDGS RQPTPLELEI ANLQGKSFYE
YVARVKW

The Phylogenetic Profile
Organisms
Enzymes